Talk given by Alex Pouget of the Department of Brain and Cognitive Sciences, at the University of Rochester. Given on May 6, 2009 to the Redwood Center for Theoretical Neuroscience at UC Berkeley.
A wide range of behaviors can be formalized as instances of probabilistic inferences. This includes odor recognition in insects, navigation in rodents, auditory localization in barn owls, decision making in primates and causal reasoning in humans, to name just a few. In all cases, the probabilistic inferences involve products of distributions and marginalization. We will show that, given the type of variability reported in neural responses, products of distributions can be implemented through linear operations over firing rates, while marginalization requires a particular nonlinearity known as quadratic divisive normalization. Both operations are conspicuous in many neural circuits raising the possibility that seemingly unrelated behaviors could in fact rely on very similar neural mechanisms across different species.